Today we reached an important milestone. Over 25% of applications in Fedora now ship AppData files. The actual numbers look like this:
- Applications with descriptions: 262/1037 (25.3%)
- Applications with keywords: 112/1037 (10.8%)
- Applications with screenshots: 235/1037 (22.7%)
- Applications in GNOME with AppData: 91/134 (67.9%)
- Applications in KDE with AppData: 5/67 (7.5%)
- Applications in XFCE with AppData: 2/20 (10.0%)
- Application addons with MetaInfo: 30
We’ve gone up a couple of percentage points in the last few weeks, mostely from the help of Ryan Lerch, who’s actually been writing AppData files and taking screenshots for upstream projects. He’s been concentrating on the developer tools for the last week or so, as this is one of the key groups of people we’re targetting for Fedora 21.
One of the things that AppData files allow us to do is be smarter suggesting “Picks” on the overview page. For 3.10 and 3.12 we had a farly short static list that we chose from at random. For 3.14 we’ve got a new algorithm that tries to find similar software to the apps you already have installed, and also suggests those. So if I have Anjunta and Devhelp installed, it might suggest D-Feet or Glade.
What exactly do you count and how is this supposed to work? The Bluefish source includes an appdata file, and it does install that file in the proper location if you run ‘make install’. But if you do not install it, the appdata file is not installed, so the package management tool cannot use it? There seems to be a chicken-egg problem here, can you help me understand this?
The AppStream metadata is generated from the whole distro archive, and shipped as a 2MB xml file.